Assessing measurement invariance of Pisa 2012 reading literacy scale among the countries determined in accordance with the language of application

The aim of the study is to examine the measurement invariance of PISA 2012 reading skills test among different countries. When determining countries, the language of scale was taken into consideration. In this study, the measurement invariance between the countries using the same language and the countries using different languages were examined respectively. Before the implementation of the PISA, the scales were first developed in English and French and the same language forms are used in different countries, therefore, firstly, countries where booklets in English and French are used are included in this study. The countries that use the English and French languages have been examined separately among themselves and the countries where English is used are determined as America, Australia, England, Ireland, Canada, Singapore and New Zealand, and the countries where French is used are determined as Belgium, France, Switzerland, Canada and Luxembourg. With these groups included in the research, the target is to determine whether there is measurement invariance between countries where the same language is used. Another objective is to examine cultures using different languages; within this scope the Turkish booklet and the booklets of the most widely spoken languages worldwide (Mandarin, Spanish, English) were compared. Representing each language, the country where the booklet number 2 is applied and where the particular language is used at most is chosen and as a result; in the study, America, Spain, Taiwan-China and Turkey sample were used. In this respect, the general purpose of the study is testing the equality of factor structure of PISA 2012 reading comprehension scale between the countries using the same or different languages. The study was carried out with 4390 student data for the group of English speaking countries, 1210 student group for the French speaking group and 2668 student data for the group speaking different languages and 28 items used in the reading skill scale in booklet 2. In the first stage, the structure of this model consisting of 28 items was confirmed by CFA (Confirmatory Factor Analysis) in all groups, and then, in order to measure whether or not inter-group measurement invariance was achieved, Multi-Group Confirmatory Factor Analysis was performed. As a result of the analysis, the invariance models created in stages were examined and different results were obtained for different country groups. It was found that structural factorial invariance was provided among the groups where English was used, but measurement invariance was not established because model 2 and model 3 models performed better. The groups using the French form are similar only in terms of factor structures and since the model 2 performed better, the measurement invariance could not be confirmed. Finally, in the analysis comparing the groups using different languages, it was found that weak factorial invariance was established, but due to the better performance of model 3, there was no measurement invariance between these groups and the comparisons between these groups were not significant. It has been demonstrated that PISA 2012 achievement scores of the reading skill scale are incomparable among the countries mentioned and that the differences may be caused by various biases rather than student performance. Keywords: Measurement invariance, confirmatory factor analysis, multi-group confirmatory factor analysis